Home
Creators
Search Recent Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

1-Click INSTALL FRAMEPACK IMAGE-TO-VIDEO WebUI! (Patreon)

Published:
2025-04-19 18:06:25
Edited:
2025-04-20 20:12:11
Imported:
2026-06
Tags:

Downloads

Content

!!! EDIT 20 APRIL: JUST UPLOADED V2 OF THE INSTALLER THAT NOW SHOULD SUPPORT 5000 SERIES GPU AND OLDER 2000/1000 SERIES AS WELL. JUST SELECT YOUR GPU SERIES AT THE BEGINNING AND THE INSTALLATION WILL DO ITS THING. BUT IF YOU SELECT OLDER GPU DO NOT CLICK ON THE UPDATE.BAT FILE AND JUST RUN THE RUN.BAT ONCE IT'S FINISHED INSTALLING!!! + ADDITION OF THE RUNPOD V2Hey everyone! I've created a 1-click installer for the newest AI model tool called "FramePack" an amazing image to video webui that can generate up to 2 minutes of videos from a single image using only 6 gb of vram! It's just amazing!You can check out the video right here: https://youtu.be/GywyMij88rYThe installer of course automates the entire install process for the webui!

LOCAL INSTALL!!!You should have already installed the 12.6 CUDA TOOLKIT from here: https://developer.nvidia.com/cuda-12-6-0-download-archive or 12.8!!!1. Download the FRAMEPACK-WEBUI_AUTO_INSTALL.bat

2. Run the bat file

3. Wait for the install to finish

4. ???5. Profit 😎

IF YOU ARE USING RUNPOD:

1. Create an account if you haven't already: Runpod

2. Click on Pod (on the left side) then click deploy

3. Choose a GPU with at least 24gb of VRAM (a cheap 4090 or an A30 are great), then click modify template and choose the pytorch 2.1 template, then edit the template and choose 80gb for both the container and volume disk, then deploy on demand

4. Go to my pods, wait for everything to finish and then click "connect", then "Connect to Jupyter Lab [Port 8888]"

5. Then drag and drop the FRAMEPACK-WEBUI_AUTO_INSTALL-RUNPOD.sh file on the left side of the UI then click on the "Terminal" icon on the right side on the UI

6. Copy and paste these two lines then press enter:

chmod +x FRAMEPACK-WEBUI_AUTO_INSTALL-RUNPOD-V2.sh./FRAMEPACK-WEBUI_AUTO_INSTALL-RUNPOD-V2.sh

7. Wait for everything to be installed

8. Click on the public url and have fun😎

IF YOU NEED TO STOP THE POD AND THEN RESTART IT. GO INSIDE THE FRAMEPACK FOLDER, CLICK ON THE TERMINAL ICON AND PASTE THESE TWO LINES THEN PRESS ENTER:

source venv/bin/activatepython demo_gradio.py --shareAs always, supporting me on Patreon allows me to keep creating helpful resources like this for the community. Thank you for your support - now go have some fun😉!

Files

Comments

哈哈 实在

Thanks. Keep up the good work but dont overwork at the same time. Take care buddy

cool1

Is the video resolution fixed or will we be able to select a specific resolution for the videos?

SPK

Bummed...will need to figure out how to set this up for pytorch sm_120. More and more of us will prob have 50xx cards going forward, if there's any way to add compat for these it would be much appreciated! We were able to figure it out with ComfyUI so I figure there's a way... PS: Super glad you're back, your channel was my gateway to this conent and has taught me a ton. Take it easy :)

DragonSin

from huggingface_hub import HfApi, ImageClassificationOutputElement, InferenceClient ImportError: cannot import name 'ImageClassificationOutputElement' from 'huggingface_hub' (C:\Users\Pc\AppData\Roaming\Python\Python310\site-packages\huggingface_hub\__init__.py) error when running run.bat after or before update.bat what can i do ? Edit: reinstalled new hugging face, now i have a new issue AttributeError: module 'torch.library' has no attribute 'register_fake' Edit2: yea ok i installed 12.8 version of cuda, and installed comfyui, that thing installed all the dependencies and now it "works"

Tomas Rampas

Awesome! I tried to give it a shot, but when I click Run it throws an error ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject . I have Cuda version 12.8. And followed all steps. Any idea how to fix this? Thx

Filip Šenjug

I had some issues with python gradio or something idk. I fixed it by upgrading diffusers and transformers Open CMD (no need for admin rights) and use this lines: pip install --upgrade diffusers After it successfully upgrades it then use this line: pip install --upgrade transformers

Killy_Blame

Traceback (most recent call last): File "G:\Ai\framepack_cu126_torch26\webui\demo_gradio.py", line 7, in import gradio as gr File "G:\Ai\framepack_cu126_torch26\system\python\lib\site-packages\gradio\__init__.py", line 3, in import gradio._simple_templates File "G:\Ai\framepack_cu126_torch26\system\python\lib\site-packages\gradio\_simple_templates\__init__.py", line 1, in from .simpledropdown import SimpleDropdown File "G:\Ai\framepack_cu126_torch26\system\python\lib\site-packages\gradio\_simple_templates\simpledropdown.py", line 7, in from gradio.components.base import Component, FormComponent File "G:\Ai\framepack_cu126_torch26\system\python\lib\site-packages\gradio\components\__init__.py", line 41, in from gradio.components.native_plot import BarPlot, LinePlot, NativePlot, ScatterPlot File "G:\Ai\framepack_cu126_torch26\system\python\lib\site-packages\gradio\components\native_plot.py", line 12, in import pandas as pd File "G:\Ai\framepack_cu126_torch26\system\python\lib\site-packages\pandas\__init__.py", line 62, in from pandas.core.api import ( File "G:\Ai\framepack_cu126_torch26\system\python\lib\site-packages\pandas\core\api.py", line 1, in from pandas._libs import ( File "G:\Ai\framepack_cu126_torch26\system\python\lib\site-packages\pandas\_libs\__init__.py", line 18, in from pandas._libs.interval import Interval File "interval.pyx", line 1, in init pandas._libs.interval ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

Zalaya123

Oof, seems not to work on 50 series cards.

Moxxi

same here, and same for the 1-Click INSTALL WAN IMAGE-TO-VIDEO & TEXT-TO-VIDEO ComfyUI WebUI! :(

Aitrepreneur

Yeah 5000 series gpu are a pain in the ass since they only work with pytorch nightly and cuda 12.8. I don't have one so I can't try it beforehand, but try this: First install cuda 12.8: https://developer.nvidia.com/cuda-downloads (restart the pc and then you can open a cmd window and check that the correct version was installed using the nvcc --version command, it should say Build cuda_12.8.r12.8/compiler.35583870_0 then launch the installer, wait for everything to be installed. Then go inside the framepack_cu126_torch26\system\python folder, click on the folder path, type cmd press enter, this will open a cmd window inside the python folder and inside type these commands: python.exe -m pip uninstall torch torchvision torchaudio python.exe -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128 then once it's done installing, go back to the framepack_cu126_torch26 folder, run the update.bat file then the run.bat file to launch the webui. If that works, I'll update the installer or do one for 5000 series GPU so that everyone can use it

Aitrepreneur

Yeah 5000 series gpu are a pain in the ass since they only work with pytorch nightly and cuda 12.8. I don't have one so I can't try it beforehand, but try this: First install cuda 12.8: https://developer.nvidia.com/cuda-downloads (restart the pc and then you can open a cmd window and check that the correct version was installed using the nvcc --version command, it should say Build cuda_12.8.r12.8/compiler.35583870_0 then launch the installer, wait for everything to be installed. Then go inside the framepack_cu126_torch26\system\python folder, click on the folder path, type cmd press enter, this will open a cmd window inside the python folder and inside type these commands: python.exe -m pip uninstall torch torchvision torchaudio python.exe -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128 then once it's done installing, go back to the framepack_cu126_torch26 folder, run the update.bat file then the run.bat file to launch the webui. If that works, I'll update the installer or do one for 5000 series GPU so that everyone can use it

Aitrepreneur

it's fixed but you can use the upscaling + interpolation workflow in comfyui that I showed in my previous video

Tomas Rampas

Do you mean on Patreon or Discord?

Aitrepreneur

Patreon yeah so that i can easily follow the convo

Filip Šenjug

But it is not working. It says wrong CUDA architecture. I am using rtx 2070 super. I installed CUDA 12.6 as mentioned and followed YT tutotial. Wont work

Aitrepreneur

Ah 2000 series, yeah ask the creator why he didn't support those GPU.... I know there is a special fork of the webui for 2000 series that might work, dm me and we can try setting it up, I can't try it on my side either. If that works, I'll make an updated installer for others as well

Filip Šenjug

Just laid down in bed as it is getting late. I tried to change some .py files to include sm75 and somehow almost got it working. Now last error is "CUDA out of memory. Tried to allocate 29.84 GiB" and of course, it just stops generating anything after that.

Jonathan Prine

I'm running 'FRAMEPACK-WEBUI_AUTO_INSTALL.bat', then 'update.bat', and 'run.bat' and keep saying "Xformers is not installed! Flash Attn is not installed! Sage Attn is installed!" I'm running a clean windows 11 install. And I installed CUDA and Python

Aitrepreneur

yup, that's normal, that means everything was installed correctly, sage attn is what needs to be installed, all the others are inferior optimization tech

GenTsoCkn

during installation, it's seems to detect CUDA version 12.8. then gives a "All components were downloaded and installed with CUDA 12.8 support" although I do have 12.9 installed, so shrug...

Brian Monarch

I noticed in the command box when you run the video creation process it says "Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB" I have a lot more VRAM than that. Is it only using 6GB? I believe in your video you mentioned that it would automatically take advantage of all your VRAM. Wasn't sure if it was after seeing that. Any suggestions on how to make it use more if it isn't already? Thanks!!!

SPK

Don't think it works still. Output after full reinstall and following these steps + start generation (default settings, vid length 5 sec) Think I see some compat errors in there... Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete. Loaded CLIPTextModel to cuda:0 as complete. Unloaded CLIPTextModel as complete. Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Loaded SiglipVisionModel to cuda:0 as complete. latent_padding_size = 27, is_last_section = False Unloaded SiglipVisionModel as complete. Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB 0%| | 0/25 [00:00

Aitrepreneur

for now there is no error at least, just need to wait

Clint

This seem to require an internet connection, if I try to run it without, it fail correction, this now works he gave me a file to replace and it now works entirely off-line

Aitrepreneur

you need internet when you start it but once the webui opens in the browser you can cut the internet and it will work

Clint

Holy crap why the file is already downloaded right? Is there any way to fix this so that you don’t need to start with the Internet?

Tom Baron

I got this error after only a minuet of starting. Anyone have an idea what the issue might be? I did upgrade to the latest Cuda. I didn't have Visual Studio installed so the installation said some parts of the CUDA might not work. Is that the likely cause? "File "G:\AI_Projects\framepack_cu126_torch26\system\python\lib\site-packages\sageattention\core.py", line 138, in sageattn raise ValueError(f"Unsupported CUDA architecture: {arch}") ValueError: Unsupported CUDA architecture: sm75"

E Farts

Tried that method, but it still didn't work Copied the last section of my logs. Not an expert on this stuff: File "E:\AI\Frame to Video\framepack_cu126_torch26\webui\diffusers_helper\models\hunyuan_video_packed.py", line 172, in __call__ hidden_states = attn_varlen_func(query, key, value, cu_seqlens_q, cu_seqlens_kv, max_seqlen_q, max_seqlen_kv) File "E:\AI\Frame to Video\framepack_cu126_torch26\webui\diffusers_helper\models\hunyuan_video_packed.py", line 111, in attn_varlen_func x = sageattn(q, k, v, tensor_layout='NHD') File "E:\AI\Frame to Video\framepack_cu126_torch26\system\python\lib\site-packages\sageattention\core.py", line 136, in sageattn return sageattn_qk_int8_pv_fp8_cuda(q, k, v, tensor_layout=tensor_layout, is_causal=is_causal, qk_quant_gran="per_warp", sm_scale=sm_scale, return_lse=return_lse, pv_accum_dtype="fp32") # sm120 has accurate fp32 accumulator for fp8 mma and triton kernel is currently not usable on sm120. File "E:\AI\Frame to Video\framepack_cu126_torch26\system\python\lib\site-packages\torch\_dynamo\eval_frame.py", line 856, in _fn return fn(*args, **kwargs) File "E:\AI\Frame to Video\framepack_cu126_torch26\system\python\lib\site-packages\sageattention\core.py", line 666, in sageattn_qk_int8_pv_fp8_cuda assert SM89_ENABLED, "SM89 kernel is not available. Make sure you GPUs with compute capability 8.9." AssertionError: SM89 kernel is not available. Make sure you GPUs with compute capability 8.9. Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete.

TheSignalNick

Mine just doesn't generate anything. Like it reads the prompt goes through the steps and then resets itself with no video

Peruvian Dota

does your resources work with 5000's gpus?

NateMac00

saw that too, but when i generated it says it was using 22.3GB or 24GB VRAM. just look at your task manager Performance

Aitrepreneur

It does, just need to so some manipulation for now, il will soon update the installer so everything is done automatically

Aitrepreneur

Don't worry about that it's not the amount of vram it uses, as NateMac00 said you can just look at the task manager to see how much vram it actually uses

Ruiz Lihuel Alejandro

Hi! This is my error (GPU RTX 3090, Windows 11) Cuda already installed, updated, this comes from run.bat: "Traceback (most recent call last): File "E:\Comfy3\FramePack2minImgVideo\framepack_cu126_torch26\webui\demo_gradio.py", line 7, in import gradio as gr File "E:\Comfy3\FramePack2minImgVideo\framepack_cu126_torch26\system\python\lib\site-packages\gradio\__init__.py", line 3, in import gradio._simple_templates File "E:\Comfy3\FramePack2minImgVideo\framepack_cu126_torch26\system\python\lib\site-packages\gradio\_simple_templates\__init__.py", line 1, in from .simpledropdown import SimpleDropdown File "E:\Comfy3\FramePack2minImgVideo\framepack_cu126_torch26\system\python\lib\site-packages\gradio\_simple_templates\simpledropdown.py", line 7, in from gradio.components.base import Component, FormComponent ... File "E:\Comfy3\FramePack2minImgVideo\framepack_cu126_torch26\system\python\lib\site-packages\pydantic\_internal\_schema_generation_shared.py", line 83, in __call__ schema = self._handler(source_type) File "E:\Comfy3\FramePack2minImgVideo\framepack_cu126_torch26\system\python\lib\site-packages\pydantic\_internal\_generate_schema.py", line 1972, in new_handler schema = metadata_get_schema(source, get_inner_schema) File "E:\Comfy3\FramePack2minImgVideo\framepack_cu126_torch26\system\python\lib\site-packages\pydantic\_internal\_std_types_schema.py", line 327, in __get_pydantic_core_schema__ constrained_schema = core_schema.list_schema(items_schema, **metadata) TypeError: list_schema() got an unexpected keyword argument 'fail_fast' Press any key to continue . . ." Edit: Thanks in advance & thank you for all of your hard work!

Passion Opera

Ran .bat file and it says it's installed but no gui shows up. How do I get to the public URL?

Zalaya123

Also didn't work for me unfortunately.

Passion Opera

Getting this error message: 'environment.bat' is not recognized as an internal or external command, operable program or batch file. The system cannot find the path specified. Press any key to continue . . . Python installed/win 11/4060ti

Killy_Blame

i got solve for me after i tried the use the other solution you mention for CUDA 12.8 custom installation. Then it started the download for the Models.

Big Ron Studd

Downloaded as the instructions suggested, everything worked as expected. Ran my first video and it worked fine. Took a little longer than expected, but didn't have any issues with the results. I've also used the GGUF img2vid in ComfyUI and it works, but the resultant video isn't always very clear or follow the prompts. I hope this one works better.

Brian Monarch

Gotcha. I noticed that if I input an image that is 1080x1920 it outputs a video that is 480x832. Is that just how it is, or is there a possibility of higher resolution on this? I just installed this on ComfyUI and there's a resolution node set to 640. If I try to make it 720 or 1080 it won't even run. I have a gpu with 48gb vram too. I'm guessing these local solutions just aren't near as good as Kling etc in terms of resolution? If there's a way to go higher that you're aware of, please let me know :) Thanks!

Brian Monarch

I noticed that if I input an image that is 1080x1920 it outputs a video that is 480x832. Is that just how it is, or is there a possibility of higher resolution on this? I just installed this on ComfyUI and there's a resolution node set to 640 default. If I try to make it 720 or 1080 it won't even run. I have a pretty beefy Nvidia A6000 gpu with 48gb vram too. I'm guessing these local solutions just aren't near as good as Kling etc in terms of resolution? If there's a way to go higher that you're aware of, please let me know :) Thanks!

Big Ron Studd

Is there anyway to tweak it to use more vram, so maybe the generation would be faster?

Dat Nguyen

anyone else getting black screens? it runs, generates previews... but eventaully all goes black.

Aerys

How do i fix this error? ValueError: Unsupported CUDA architecture: sm75

555 Londus

5090 don t work need sm120, or?

Claus Schober

I have the same error. I use ShadowPC with a RTX A6000 Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Loaded SiglipVisionModel to cuda:0 as complete. latent_padding_size = 27, is_last_section = False Unloaded SiglipVisionModel as complete. Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB 0%| | 0/25 [00:00

John Cabrer

Same error for me RTX2070 Super. ValueError: Unsupported CUDA architecture: sm75

John Cabrer

means that SageAttention, which your video generation pipeline is using for optimized attention, does not support your GPU architecture — specifically, the sm_75 architecture. 🔍 What is sm_75? sm_75 corresponds to NVIDIA Turing architecture GPUs like: RTX 2060 / 2070 / 2080 (Super, Ti, etc.) Quadro RTX 3000/4000/5000 You have a few ways to proceed, depending on your priorities: Option 1: Disable SageAttention (Recommended for Compatibility) Use a more compatible attention backend (like FlashAttention, if available, or default PyTorch attention). In your command or script, disable SageAttention explicitly. Look for a config or launch flag that says something like: bash Copy Edit --no-sage-attn Or edit the config file if applicable and set: python Copy Edit "attention_backend": "mem_efficient" # or "flash" if supported Option 2: Modify SageAttention to Allow SM75 (Risky) This would require modifying the source of sageattention to bypass the architecture check and recompiling the extension manually with TORCH_CUDA_ARCH_LIST=7.5. Not ideal unless you’re experienced with C++/CUDA build pipelines and understand the risk (it might still not work or crash later). Option 3: Upgrade Your GPU Only consider this if you're heavily invested in SageAttention and want the fastest performance. GPUs with sm_80 or higher (e.g., RTX 30-series, 40-series, A100, etc.) are supported. 🧠 Quick Summary Option Description Effort Safety 🔧 Disable SageAttention Most compatible fix Low ✅ Safe ⚙️ Patch + Recompile SageAttention Advanced workaround High ❌ Risky 💸 Upgrade GPU Best for performance $$$ ✅ Safe (if feasible)

Ruiz Lihuel Alejandro

Solved: Open terminal here: framepack_cu126_torch26\system\python Then: python.exe -m pip install --upgrade pydantic fastapi And finally: python.exe -m pip install --upgrade gradio

Virtamouse

Works great! Thanks! New ultimatecomfy workflow when? I need all the new tools!!

Instruments of Destruction

raise ValueError(f"Unsupported CUDA architecture: {arch}") ValueError: Unsupported CUDA architecture: sm75

Instruments of Destruction

Same error for me: raise ValueError(f"Unsupported CUDA architecture: {arch}") ValueError: Unsupported CUDA architecture: sm75

Aitrepreneur

I just uploaded the V2 installer, should work now. Just select your gpu series at the beginning and make sure you have cuda 12.8 installed

Aitrepreneur

you can use the comfyui workflow for upscaling and video interpolation that I showed in my last wan video for anything like that, much better and faster

Aitrepreneur

it should open in the brower automatically once you run the run.bat file

Aitrepreneur

you can use the comfyui workflow for upscaling and video interpolation that I showed in my last wan video for anything like that, much better and faster

Aitrepreneur

what's your GPU? Have you selected 16 for mp4 compression parameter?

Aitrepreneur

I just uploaded the V2 installer, should work now. Just select your gpu series at the beginning and make sure you have cuda 12.6 installed and do not update it

Aitrepreneur

I just uploaded the V2 installer, should work now. Just select your gpu series at the beginning and make sure you have cuda 12.8 installed

Aitrepreneur

I just uploaded the V2 installer, should work now. Just select your gpu series at the beginning and make sure you have cuda 12.6 installed and do not update it

555 Londus

And when i use Runpod? will there be also a V2 version?

Aitrepreneur

No need runpod should work right out of the box

555 Londus

It hasn t i test it with 5090 :) , 4090 works A40 works but with 5090 it won t. ^^ But thx for your quick answer.

Aitrepreneur

Aaaah OK yeah, I'll look into it but you can still rent a 4090 for now

Junbi

V2 installer working on 5080. Thank you!

Passion Opera

Why do I get this when I run the run.bat file? The program is loaded in a writable directory and I've tried it in two different directories: There was a problem when trying to write in your cache folder (C:\Windows\System32\framepack_cu126_torch26\webui\hf_download\hub). You should set the environment variable TRANSFORMERS_CACHE to a writable directory. Currently enabled native sdp backends: ['flash', 'math', 'mem_efficient', 'cudnn'] Xformers is not installed! Flash Attn is not installed! Sage Attn is installed! Namespace(share=False, server='127.0.0.1', port=None, inbrowser=True) Free VRAM 6.8994140625 GB High-VRAM Mode: False Traceback (most recent call last): File "C:\Windows\System32\framepack_cu126_torch26\webui\demo_gradio.py", line 49, in text_encoder = LlamaModel.from_pretrained("hunyuanvideo-community/HunyuanVideo", subfolder='text_encoder', torch_dtype=torch.float16).cpu() File "C:\Windows\System32\framepack_cu126_torch26\system\python\lib\site-packages\transformers\modeling_utils.py", line 3506, in from_pretrained resolved_config_file = cached_file( File "C:\Windows\System32\framepack_cu126_torch26\system\python\lib\site-packages\transformers\utils\hub.py", line 403, in cached_file resolved_file = hf_hub_download( File "C:\Windows\System32\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\utils\_validators.py", line 114, in _inner_fn return fn(*args, **kwargs) File "C:\Windows\System32\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\file_download.py", line 961, in hf_hub_download return _hf_hub_download_to_cache_dir( File "C:\Windows\System32\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\file_download.py", line 1078, in _hf_hub_download_to_cache_dir os.makedirs(os.path.dirname(blob_path), exist_ok=True) File "os.py", line 215, in makedirs File "os.py", line 215, in makedirs File "os.py", line 215, in makedirs File "os.py", line 225, in makedirs PermissionError: [WinError 5] Access is denied: 'C:\\Windows\\System32\\framepack_cu126_torch26\\webui\\hf_download'

Passion Opera

I did not run as admin, as you told me not to earlier. How do I run the bat files from the cmd prompt without getting all these errors? All I'm trying to do us get the gui to open and it doesn't work no matter what drive the program is on.

Aitrepreneur

OK I added the V2 installer for Runpod as well, it will work with 5090 GPU as well and is twice as fast compared to the V1, just select the correct GPU during the install

Aitrepreneur

but your install is in system32.....meaning that you did at one point run the installer as admin...

MOG. SEOUL

The installation is complete and the GUI window appears, but there is no video output. To create a public link, set `share=True` in `launch()`. Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete. Loaded CLIPTextModel to cuda:0 as complete. Unloaded CLIPTextModel as complete. Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Loaded SiglipVisionModel to cuda:0 as complete. latent_padding_size = 27, is_last_section = False Unloaded SiglipVisionModel as complete. Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB 0%| | 0/25 [00:00 return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs) File "C:\Users\drunk\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\triton\runtime\jit.py", line 653, in run kernel.run(grid_0, grid_1, grid_2, stream, kernel.function, kernel.packed_metadata, launch_metadata, File "C:\Users\drunk\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\triton\compiler\compiler.py", line 399, in __getattribute__ self._init_handles() File "C:\Users\drunk\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\triton\compiler\compiler.py", line 394, in _init_handles self.module, self.function, self.n_regs, self.n_spills = driver.active.utils.load_binary( SystemError: PY_SSIZE_T_CLEAN macro must be defined for '#' formats Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete.

Passion Opera

I see. But is there some way to do a clean install of this app? I've deleted and re-installed several times and keep running into the same issue. I guess I'll just try the new V2 and see if that works.

Aitrepreneur

Yeah just run the bat file normally, nothing should be run as admin

anime draco

hey i have a problem with the framepack installation , i just used the bat file and downloaded everythging then i did press run.bat , but "Loading checkpoint shards" gets stuck at 2/4

Aitrepreneur

nothing you can do but wait, relaunch the run.bat file as well

Claudinei

WIN 11 Pro RTX 2060 12GB I installed the "CUDA Toolkit 12.6" I rebooted and after that I restarted the PC, I ran the script FRAMEPACK-WEBUI_AUTO_INSTALL-V2.bat I selected option 3 (Turing / Pascal / etc.) when I try to use it I get the error I tried twice and the same error happened File "D:\IA\framepack_cu126_torch26\system\python\lib\site-packages\sageattention\core.py", line 138, in sageattn raise ValueError(f"Unsupported CUDA architecture: {arch}") ValueError: Unsupported CUDA architecture: sm75 Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete.

Aitrepreneur

Are you sure you didn't run the update.bat file because if you did you need to redo the install, if not then I guess you can try this: go inside the framepack_cu126_torch26\system\python folder, click on the folder path, type cmd press enter, this will open a cmd window inside the python folder and inside type these commands: python.exe -m pip uninstall -y sageattention triton python.exe -s -m pip install triton-windows pip install sageattention==1.0.6 then relaunch the run.bat file, never run the update.bat ever

Tom Baron

So my RTX 2080TI is going to be to old to run with this correct?

Instruments of Destruction

Got it. However, the first time we ran the file and was asked to update Cuda, we did. Therefore we can't go back to an older version. I tried this V2 installer and still gives the same result raise ValueError(f"Unsupported CUDA architecture: {arch}") ValueError: Unsupported CUDA architecture: sm75

Claudinei

I redid the entire installation from scratch, without running update.bat, and the problem remains the same. Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB 0%| | 0/25 [00:00

Wing Tang Wong

Great work! Download, installed, and running.

Josen Zhong

Same for me. File "E:\AI\framepack_cu126_torch26\system\python\lib\site-packages\sageattention\core.py", line 138, in sageattn raise ValueError(f"Unsupported CUDA architecture: {arch}") ValueError: Unsupported CUDA architecture: sm75 Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete.

Virtamouse

I 2nd this. Looks very cool.

HealingPaint

Just guessing....out of disc space maybe?

E Farts

Yeah the new one did it once I restarted the process. Thank you! It's honestly really cool to see it work. You are a gentleman and a scholar!

M.K

Hi, i keep getting this error , i have rtx3060 on a laptop, and i get this error when i click start generation File "D:\framepack_cu126_torch26\system\python\lib\site-packages\triton\runtime\jit.py", line 330, in return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs) File "D:\framepack_cu126_torch26\system\python\lib\site-packages\triton\runtime\jit.py", line 623, in run kernel = self.compile( File "D:\framepack_cu126_torch26\system\python\lib\site-packages\triton\compiler\compiler.py", line 283, in compile next_module = compile_ir(module, metadata) File "D:\framepack_cu126_torch26\system\python\lib\site-packages\triton\backends\nvidia\compiler.py", line 407, in stages["cubin"] = lambda src, metadata: self.make_cubin(src, metadata, options, self.capability) File "D:\framepack_cu126_torch26\system\python\lib\site-packages\triton\backends\nvidia\compiler.py", line 389, in make_cubin raise RuntimeError(f'{error}\n' RuntimeError: `ptxas` failed with error code 4294967295 `ptxas` stderr: ptxas fatal : Memory allocation failure ptxas fatal : Ptx assembly aborted due to errors Repro command: D:\framepack_cu126_torch26\system\python\lib\site-packages\triton\backends\nvidia\bin\ptxas.exe -lineinfo -v --gpu-name=sm_86 C:\TEMP\tmpdiiundm9.ptx -o C:\TEMP\tmpdiiundm9.ptx.o

Aerys

I'm also getting this error, did you manage to fix it?

Aerys

Using RTX 2070: Using V2 Installer, selected option 3, with Cuda 12.6 installed. Getting error: raise ValueError(f"Unsupported CUDA architecture: {arch}") ValueError: Unsupported CUDA architecture: sm75 I haven't ran Update.bat . Tried reinstalling but same error.

HealingPaint

Glad to hear this. I thought it was broken, but it does seem to work. On a RTX-4090 should it be taking roughly a minute per 25 steps (1min) generated for each video?

Claudinei

So far I haven't been able to solve it, I've tried 3 times to reinstall everything from scratch

Jason

"Sage Attn is not installed!" even tried deleting and redoing the install. I dont see errors on install and it actually does load and seems to work but i assume under lower quality settings/tools? I run the environment.bat to run pip and install but i get "Unable to create process using '"D:\webui-forge\system\python\python.exe" " and maybe thats why?

Nikki Newsome (edited)

Comment edits

2026-09-08 20:00 (Win 11, 3060) Got it all to run successfully, 25/25 Sampling completed, however... Gets to 100%, then errors. I can't seem to find any outputs (other than a PNG) in explorer. "To create a public link, set `share=True` in `launch()`. Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete. Loaded CLIPTextModel to cuda:0 as complete. Unloaded CLIPTextModel as complete. Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Loaded SiglipVisionModel to cuda:0 as complete. latent_padding_size = 9, is_last_section = False Unloaded SiglipVisionModel as complete. Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB 100%|██████████████████████████████████████████████████████████████████████████████████| 25/25 [12:19<00:00, 29.58s/it] Offloading DynamicSwap_HunyuanVideoTransformer3DModelPacked from cuda:0 to preserve memory: 8 GB Press any key to continue . . ."
2025-04-21 09:37

TAGILUX

Using the V2 installer and I have a RTX5000 GPU - I installed - tried with both update.bat and without - got the same error both times Traceback (most recent call last): File "C:\development\framepack_video_ai\framepack_cu126_torch26\webui\demo_gradio.py", line 17, in from diffusers import AutoencoderKLHunyuanVideo ImportError: cannot import name 'AutoencoderKLHunyuanVideo' from 'diffusers' (C:\Users\admin\AppData\Roaming\Python\Python310\site-packages\diffusers\__init__.py) Press any key to continue . . .

Jean Dupont

Happy to see you back and great news and work! Simple question, your bat file, does it still work if i have cuda 12.8 and not specificly the 12.6? I think i heard you say you have 12.8 and it's ok. But doesn it work with the bat file?

earlmckay

Thanks, it worked straight away. But the Promt realization does not work well for me - or is there a switch for NSFW that you have to activate first?

MOG. SEOUL

4070 ti and Build cuda_12.6.r12.6

Passion Opera

It works! Thank you. I was impatient after running the bat file. I expected an instant startup like my other gradio apps. Left the black screen open all night and the gradio finally appeared.

Passion Opera

I had the same issue. It won't open right away because it's downloading big models on first open. I left my black window open all night and the gradio gui finally appeared.

John Cabrer

RTX 2070 TI. CUDA 12.6. sm75 value error.

Elkin Marenco

I started with the same error (I also didn't run the update.bat) Then I applied the solution from Aitrepreneur python.exe -m pip uninstall -y sageattention triton python.exe -s -m pip install triton-windows pip install sageattention==1.0.6 and apparently everything was ok But when I ran it with the default values for a video of 5 secconds in my RTX 2060, I got this Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete. Loaded CLIPTextModel to cuda:0 as complete. Unloaded CLIPTextModel as complete. Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Loaded SiglipVisionModel to cuda:0 as complete. latent_padding_size = 27, is_last_section = False Unloaded SiglipVisionModel as complete. Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB 100%|██████████████████████████████████████████████████████████████████████████████████| 25/25 [18:19<00:00, 43.96s/it] Offloading DynamicSwap_HunyuanVideoTransformer3DModelPacked from cuda:0 to preserve memory: 8 GB Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Decoded. Current latent shape torch.Size([1, 16, 9, 80, 76]); pixel shape torch.Size([1, 3, 33, 640, 608]) latent_padding_size = 18, is_last_section = False Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB Exception in callback _ProactorBasePipeTransport._call_connection_lost(None) handle: Traceback (most recent call last): File "asyncio\events.py", line 80, in _run File "asyncio\proactor_events.py", line 162, in _call_connection_lost ConnectionResetError: [WinError 10054] Se ha forzado la interrupción de una conexión existente por el host remoto 100%|██████████████████████████████████████████████████████████████████████████████████| 25/25 [18:58<00:00, 45.54s/it] Offloading DynamicSwap_HunyuanVideoTransformer3DModelPacked from cuda:0 to preserve memory: 8 GB Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Decoded. Current latent shape torch.Size([1, 16, 18, 80, 76]); pixel shape torch.Size([1, 3, 69, 640, 608]) latent_padding_size = 9, is_last_section = False Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB 0%| | 0/25 [00:00 Traceback (most recent call last): File "asyncio\events.py", line 80, in _run File "asyncio\proactor_events.py", line 162, in _call_connection_lost ConnectionResetError: [WinError 10054] Se ha forzado la interrupción de una conexión existente por el host remoto 84%|████████████████████████████████████████████████████████████████████▉ | 21/25 [14:25<03:09, 47.26s/it] The last one never ends and never finished and the result is a 2 seconds all black video. Update After 20 more minutes started a fourth attempt Offloading DynamicSwap_HunyuanVideoTransformer3DModelPacked from cuda:0 to preserve memory: 8 GB Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Decoded. Current latent shape torch.Size([1, 16, 27, 80, 76]); pixel shape torch.Size([1, 3, 105, 640, 608]) latent_padding_size = 0, is_last_section = True Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB Exception in callback _ProactorBasePipeTransport._call_connection_lost(None) handle: Traceback (most recent call last): File "asyncio\events.py", line 80, in _run File "asyncio\proactor_events.py", line 162, in _call_connection_lost ConnectionResetError: [WinError 10054] Se ha forzado la interrupción de una conexión existente por el host remoto Exception in callback _ProactorBasePipeTransport._call_connection_lost(None) handle: Traceback (most recent call last): File "asyncio\events.py", line 80, in _run File "asyncio\proactor_events.py", line 162, in _call_connection_lost ConnectionResetError: [WinError 10054] Se ha forzado la interrupción de una conexión existente por el host remoto 0%| | 0/25 [00:00

Bill P

im using a 5080 seems to work so far but how to modify so it installs where i want not in the downloads folder thanks

Jim Moore

Ok so I've tried to install this with no luck. Windows 11 RTX 4070 When I run the .bat file I just keep getting this error: File "C:\AI\framepack_cu126_torch26\system\python\lib\site-packages\diffusers\utils\import_utils.py", line 822, in _get_module raise RuntimeError( RuntimeError: Failed to import diffusers.loaders.single_file_model because of the following error (look up to see its traceback): Failed to import diffusers.schedulers.scheduling_dpmsolver_multistep because of the following error (look up to see its traceback): numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\AI\framepack_cu126_torch26\webui\demo_gradio.py", line 17, in from diffusers import AutoencoderKLHunyuanVideo File "", line 1075, in _handle_fromlist File "C:\AI\framepack_cu126_torch26\system\python\lib\site-packages\diffusers\utils\import_utils.py", line 811, in __getattr__ value = getattr(module, name) File "C:\AI\framepack_cu126_torch26\system\python\lib\site-packages\diffusers\utils\import_utils.py", line 810, in __getattr__ module = self._get_module(self._class_to_module[name]) File "C:\AI\framepack_cu126_torch26\system\python\lib\site-packages\diffusers\utils\import_utils.py", line 822, in _get_module raise RuntimeError( RuntimeError: Failed to import diffusers.models.autoencoders.autoencoder_kl_hunyuan_video because of the following error (look up to see its traceback): Failed to import diffusers.loaders.single_file_model because of the following error (look up to see its traceback): Failed to import diffusers.schedulers.scheduling_dpmsolver_multistep because of the following error (look up to see its traceback): numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject Press any key to continue . . .

Aitrepreneur

try this: go inside the framepack_cu126_torch26\system\python folder, click on the folder path, type cmd press enter, this will open a cmd window inside the python folder and inside type these commands: python.exe -m pip uninstall -y sageattention triton python.exe -s -m pip install triton-windows pip install sageattention==1.0.6

Aitrepreneur

just follow what I show in the video, why are you running environment.bat? what's your GPU?

Aitrepreneur

change the preserved vram from 6 to 9 and try again.

Aitrepreneur

no, just a good prompt, depends on what you are generating of course

Aitrepreneur

just take the entiere folder and put it where you want, it's a portable version after all

LW

Works fine for me. Nice to be able to make longer videos than just a couple of seconds. Looking forward to the ability to use LoRas to drive the videos in the desired direction.

M.K

so it cant run on 3060 laptop ?

Tom Baron

Thanks for the update, Works great.

JCS

My videos are just a video of an all black screen. What am I doing wrong?

Jason

4090... i ran environment.bat because i was troubleshooting hoping not to bother you and maybe provide some information assist when i did. I did follow the video, its very easy. download bat file.. run bat file.. run update.. run run... make money. Only for me all three times doing that, it says "Sage Attn is not installed!". 1. It installs fine (Installing collected packages: sageattention -- Successfully installed sageattention-2.1.1+cu126torch2.6.0) 2. It opens.. no errors during the install (I can share the last install command line execution). 3. I do get results with this, but I am assuming without optimal settings/tools.

Tim Eilers

* Running on local URL: http://127.0.0.1:7860 To create a public link, set `share=True` in `launch()`. Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete. Loaded CLIPTextModel to cuda:0 as complete. Unloaded CLIPTextModel as complete. Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Loaded SiglipVisionModel to cuda:0 as complete. latent_padding_size = 27, is_last_section = False Unloaded SiglipVisionModel as complete. Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB 100%|██████████████████████████████████████████████████████████████████████████████████| 25/25 [00:30<00:00, 1.23s/it] Offloading DynamicSwap_HunyuanVideoTransformer3DModelPacked from cuda:0 to preserve memory: 8 GB Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. C:\Users\timti\Desktop\framepack_cu126_torch26\system\python\lib\site-packages\torchvision\io\_video_deprecation_warning.py:5: UserWarning: The video decoding and encoding capabilities of torchvision are deprecated from version 0.22 and will be removed in version 0.24. We recommend that you migrate to TorchCodec, where we'll consolidate the future decoding/encoding capabilities of PyTorch: https://github.com/pytorch/torchcodec warnings.warn( Decoded. Current latent shape torch.Size([1, 16, 9, 104, 60]); pixel shape torch.Size([1, 3, 33, 832, 480]) latent_padding_size = 18, is_last_section = False Moving DynamicSwap_HunyuanVideoTransformer3DModelPacked to cuda:0 with preserved memory: 6 GB Traceback (most recent call last): File "C:\Users\timti\Desktop\framepack_cu126_torch26\system\python\lib\site-packages\ffmpy.py", line 95, in run self.process = subprocess.Popen( File "subprocess.py", line 969, in __init__ File "subprocess.py", line 1438, in _execute_child FileNotFoundError: [WinError 2] The system cannot find the file specified During handling of the above exception, another exception occurred: Traceback (most recent call last): File "C:\Users\timti\Desktop\framepack_cu126_torch26\system\python\lib\site-packages\gradio\queueing.py", line 715, in process_events response = await route_utils.call_process_api( File "C:\Users\timti\Desktop\framepack_cu126_torch26\system\python\lib\site-packages\gradio\route_utils.py", line 322, in call_process_api output = await app.get_blocks().process_api( File "C:\Users\timti\Desktop\framepack_cu126_torch26\system\python\lib\site-packages\gradio\blocks.py", line 2147, in process_api data = await self.postprocess_data(block_fn, result["prediction"], state) File "C:\Users\timti\Desktop\framepack_cu126_torch26\system\python\lib\site-packages\gradio\blocks.py", line 1939, in postprocess_data prediction_value = block.postprocess(prediction_value) File "C:\Users\timti\Desktop\framepack_cu126_torch26\system\python\lib\site-packages\gradio\components\video.py", line 283, in postprocess processed_files = (self._format_video(value), None) File "C:\Users\timti\Desktop\framepack_cu126_torch26\system\python\lib\site-packages\gradio\components\video.py", line 339, in _format_video and not processing_utils.video_is_playable(video) File "C:\Users\timti\Desktop\framepack_cu126_torch26\system\python\lib\site-packages\gradio\processing_utils.py", line 1104, in video_is_playable output = probe.run(stderr=subprocess.PIPE, stdout=subprocess.PIPE) File "C:\Users\timti\Desktop\framepack_cu126_torch26\system\python\lib\site-packages\ffmpy.py", line 105, in run raise FFExecutableNotFoundError( ffmpy.FFExecutableNotFoundError: Executable 'ffprobe' not found 100%|██████████████████████████████████████████████████████████████████████████████████| 25/25 [00:32<00:00, 1.31s/it] Offloading DynamicSwap_HunyuanVideoTransformer3DModelPacked from cuda:0 to preserve memory: 8 GB Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete.

Magnus Roe

I would like to install this into stability matrix, do I have to copy it over afterwards or is there a simpler way?

studiosmh

I have this same error message too.

Michael A Livingston

I have the same issue and I am following your instructions. After executing the first command I get "WARNING: Skipping triton as it is not installed." After following your instructions, it works. Thank you.

SuperBestDavid

This this this! I don't know if this is the appropriate place to request features, but I really hope LoRa support gets added quickly!

ak pak

Change mp4 compression to 16 or higher.

Lagartija

where do you make this change?

Lagartija

its taking me about 30 minutes to make a 2 second clip which renders all black with all the default settings (minus time) im on a 3080ti 16gb. any ideas? video gen has always been dead slow on my system

ant

This is awesome, is there a way to get it to run without being connected to the internet?

TAGILUX

That is the full log on loading run.bat

anime draco

00<00:00, 214.93it/s] Loading checkpoint shards: 0%| | 0/4 [00:00 text_encoder = LlamaModel.from_pretrained("hunyuanvideo-community/HunyuanVideo", subfolder='text_encoder', torch_dtype=torch.float16).cpu() File "D:\telechargements\framepack_cu126_torch26\system\python\lib\site-packages\transformers\modeling_utils.py", line 4225, in from_pretrained ) = cls._load_pretrained_model( File "D:\telechargements\framepack_cu126_torch26\system\python\lib\site-packages\transformers\modeling_utils.py", line 4706, in _load_pretrained_model state_dict = load_state_dict( File "D:\telechargements\framepack_cu126_torch26\system\python\lib\site-packages\transformers\modeling_utils.py", line 555, in load_state_dict with safe_open(checkpoint_file, framework="pt") as f: (os error 1455)

Aitrepreneur

There is already a pull request for that feature, I'm waiting until it gets fused into the main repo. I might make an updated video once it's done.

Aitrepreneur

It's a portable version so you can just move the whole folder where you want

Magnus Roe

I see, but I meant to open it as a workflow in Comfyui (within stability matrix), is it possible to convert it?

ant

Worked, thanks man!

Rococopops

Can this be used to bring anime or art come to life?

LW

Cool. AI is developing fast but we still have to be a bit patient.... while we wait for AI to realise that most humans are a waste of space and wipe out humanity... except for lolis of course... even AI knows they are too cute to wipe out!

Ryan Davies

im getting a few error messages with 'register_fake' at the end 4090. install looks to have gone through without any issues

jack2pot

im simple getting this after last step (run .bat): 'environment.bat' is not recognized as an internal or external command, operable program or batch file. The system cannot find the path specified. Press any key to continue . . .

Rococopops

I updated to Cuda 12.6, restarted, then downloaded the 1 click installer but came across something weird that I'm not sure if I should continue the install.

Rococopops

I hit enter by mistake, I get this text with jargon, I have an RTX 4000 so I know to select 1, but is the below text I see normal? ================================================================================ FRAMEPACK┬áWEBUI┬á+┬áTriton┬á+┬áSageAttention┬á2┬áInstaller ================================================================================ Which NVIDIA GPU series are you using? 1) RTX┬á3000┬á/┬á4000 (Ampere / Ada-Lovelace) 2) RTX┬á5000 (Ada-Next / ÔÇ£BlackwellÔÇØ) 3) RTX┬á2000┬áor older (Turing / Pascal / etc.) Enter the number of your GPU series (1ÔÇæ3):

Kuatoe

Getting the same issue.

JR Dom

RTX 3070 ti 6gb Win 11 Having a very slow time just to generate a 5 second video. Its able to create the 25 samples but stays forever to generate the end video. Curious how long should I expect this to render vs knowing that its just frozen or at an error to generate? this is my last line in terminal as its still rendering "Loaded AutoencoderKLHunyanVideo to cuda:0 as complete." I should also add that as im writing this post its been rendering 2 hrs prior to me posting here. btw very grateful for you creating all this!!! UPDATE: Finished rendering 5 sec in 5 hrs. Would it normally take this long?

ChubbsMcBeef

I'm having the same issue. At the end it has this error: Extraction failed ÔÇô check the archive and try again.

Trooperware

Thanks for your work. I have one question... Does computer characteristics matter or only GPU?

Testtn

I tried it on Runpod with an H100 XSM 80gb Vram... 15 second clips were a few minutes. Tried on an H200 but nominal difference.

Mol Jaa Boss

can this be done with a AMD card?

Claudinei

Do you know how to resolve this error? ptxas C:\Users\User\AppData\Local\Temp\tmpv8tc40bd.ptx, line 895; error : Feature 'cvt with .f32.bf16' requires .target sm_80 or higher ptxas fatal : Ptx assembly aborted due to errors Repro command: D:\IA\framepack_cu126_torch26\system\python\lib\site-packages\triton\backends\nvidia\bin\ptxas.exe -lineinfo -v --gpu-name=sm_75 C:\Users\User\AppData\Local\Temp\tmpv8tc40bd.ptx -o C:\Users\User\AppData\Local\Temp\tmpv8tc40bd.ptx.o Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete.

Zoltan Vegh

thank you, great work! local is working but very slow, tried runpod, deploying pods two times, 4090, even 2x 4090 with 36 gb ram, and got every time this error message, cound not connect while I was charged, not sure if I was just not lucky or their service model is like this, but it was disappointing: error creating container: container: create: container create: Error response from daemon: layer does not exist - pod was running with this error messages while clock was ticking, connection was not possible...

vrai honnête

Everything is going great. Thank you for the effort! One question, I cannot get it to work while offline. There allways seems to be a connection necessary to the hugginface servers. Can you imagine why that is? I thought that all the necessary enviornment would be 100% local but it seems that this is not the case. Greetings.

Z

Seeing the same message. But I just realized I might be using an unsupported GPU. Is that what runpod is for? Confused by some of this.

PaK

File "X:\framepack_cu126_torch26\webui\demo_gradio.py", line 216, in worker move_model_to_device_with_memory_preservation(transformer, target_device=gpu, preserved_memory_gb=gpu_memory_preservation) File "X:\framepack_cu126_torch26\webui\diffusers_helper\memory.py", line 92, in move_model_to_device_with_memory_preservation if hasattr(m, 'weight'): File "X:\framepack_cu126_torch26\webui\diffusers_helper\memory.py", line 26, in hacked_get_attr return torch.nn.Parameter(p.to(**kwargs), requires_grad=p.requires_grad) torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 108.00 MiB. GPU 0 has a total capacity of 24.00 GiB of which 15.23 GiB is free. Of the allocated memory 7.44 GiB is allocated by PyTorch, and 33.92 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables) This happens every time with my 3090 setup.

NoFuckingGame

Win 10, 4060 ti with 16 gb of vram, and i got this error. can somebody help?: To create a public link, set `share=True` in `launch()`. Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete. Loaded CLIPTextModel to cuda:0 as complete. Unloaded CLIPTextModel as complete. Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Loaded SiglipVisionModel to cuda:0 as complete. Traceback (most recent call last): File "D:\Ai video\framepack_cu126_torch26\webui\demo_gradio.py", line 180, in worker rnd = torch.Generator("cpu").manual_seed(seed) RuntimeError: manual_seed expected a long, but got NoneType Unloaded SiglipVisionModel as complete. Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete. Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete. Loaded CLIPTextModel to cuda:0 as complete. Unloaded CLIPTextModel as complete. Loaded AutoencoderKLHunyuanVideo to cuda:0 as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Loaded SiglipVisionModel to cuda:0 as complete. Traceback (most recent call last): File "D:\Ai video\framepack_cu126_torch26\webui\demo_gradio.py", line 180, in worker rnd = torch.Generator("cpu").manual_seed(seed) RuntimeError: manual_seed expected a long, but got NoneType Unloaded SiglipVisionModel as complete. Unloaded DynamicSwap_LlamaModel as complete. Unloaded CLIPTextModel as complete. Unloaded SiglipVisionModel as complete. Unloaded AutoencoderKLHunyuanVideo as complete. Unloaded DynamicSwap_HunyuanVideoTransformer3DModelPacked as complete.

Vast Universe

beautiful code on the shell scripts btw. By any change do you know if the civitai wan2.1 is the same as the one in huggingface?

Vast Universe

File "D:\Ai video\framepack_cu126_torch26\webui\demo_gradio.py", line 180, in worker rnd = torch.Generator("cpu").manual_seed(seed) RuntimeError: manual_seed expected a long, but got NoneType i guess you can just put a random number on the seed

ak pak

with RTX 3080 10GB I got 4 sec video in about 20-25 minutes.

Don

Any idea why it keeps getting stuck? It was working fine and I didn't change anything except a new image. now it gets stuck on "moving dynamicswap_hunyuanvideotransformer3dmodelpacked to cuda:0 with preserved memory: 6GB 0% and GPU pegged at 100% / max VRAM

TAGILUX

Any updates please??

Den Sh

!! self.initialize_options() Traceback (most recent call last): File "/workspace/FramePack/SageAttention/setup.py", line 175, in setup( File "/workspace/FramePack/venv/lib/python3.10/site-packages/setuptools/__init__.py", line 117, in setup return distutils.core.setup(**attrs) File "/workspace/FramePack/venv/lib/python3.10/site-packages/setuptools/_distutils/core.py", line 186, in setup return run_commands(dist) File "/workspace/FramePack/venv/lib/python3.10/site-packages/setuptools/_distutils/core.py", line 202, in run_commands dist.run_commands() File "/workspace/FramePack/venv/lib/python3.10/site-packages/setuptools/_distutils/dist.py", line 1002, in run_commands self.run_command(cmd) File "/workspace/FramePack/venv/lib/python3.10/site-packages/setuptools/dist.py", line 1104, in run_command super().run_command(command) File "/workspace/FramePack/venv/lib/python3.10/site-packages/setuptools/_distutils/dist.py", line 1021, in run_command cmd_obj.run() File "/workspace/FramePack/venv/lib/python3.10/site-packages/setuptools/command/install.py", line 105, in run self.do_egg_install() File "/workspace/FramePack/venv/lib/python3.10/site-packages/setuptools/command/install.py", line 143, in do_egg_install raise NotImplementedError("Support for egg-based install has been removed.") NotImplementedError: Support for egg-based install has been removed. root@c7ec50184876:/workspace# can't install it. an error

Vegetables Cut

I'm facing the same problem using Runpod

James Jones

When running update.bat in the framepack folder (after install and PC restart). fatal: detected dubious ownership in repository at 'C:/Windows/System32/framepack_cu126_torch26/webui'

Edgar

same problem

James Jones

Any support for this issue?

Tom Baron

Those commands stopped the sm75 error I was getting. Using A 2080 Ti

H-BOX

Hi ! thank you so much for this :) I wanted to ask, is there a way to reduce framerate ? As a 2d animator, I know that in sojme styles being at 2(12fps), or at 3 is totally fine to look at. and it would be so much faster to produce.

H-BOX

Also, I can't see the process in realtime like you do in the video, do I have a different version ? I just bought a new laptop with 5080

James Jones

Has he stopped providing support?

TAGILUX

Any ideas?

Jeremiah Everett

I'm getting the same thing as you! Did you find a resolution?

xdsp

Can this make use of the Hunyaun LoRAs available on civitai? If so, how do we use the LoRAs in the webui?

Jeremiah Everett

What is the fix if it was already run as admin? I've run it not as admin and nothing comes up. Just a black screen!

Jeremiah Everett

update. Here is the results with run not as admin. but still didn't work. Currently enabled native sdp backends: ['flash', 'math', 'mem_efficient', 'cudnn'] Xformers is not installed! Flash Attn is not installed! Sage Attn is installed! Namespace(share=False, server='127.0.0.1', port=None, inbrowser=True) Free VRAM 6.890625 GB High-VRAM Mode: False Traceback (most recent call last): File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\connection.py", line 198, in _new_conn sock = connection.create_connection( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\util\connection.py", line 85, in create_connection raise err File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\util\connection.py", line 73, in create_connection sock.connect(sa) TimeoutError: timed out The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\connectionpool.py", line 793, in urlopen response = self._make_request( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\connectionpool.py", line 491, in _make_request raise new_e File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\connectionpool.py", line 467, in _make_request self._validate_conn(conn) File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\connectionpool.py", line 1099, in _validate_conn conn.connect() File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\connection.py", line 616, in connect self.sock = sock = self._new_conn() File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\connection.py", line 207, in _new_conn raise ConnectTimeoutError( urllib3.exceptions.ConnectTimeoutError: (, 'Connection to huggingface.co timed out. (connect timeout=10)') The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\requests\adapters.py", line 486, in send resp = conn.urlopen( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\connectionpool.py", line 847, in urlopen retries = retries.increment( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\urllib3\util\retry.py", line 515, in increment raise MaxRetryError(_pool, url, reason) from reason # type: ignore[arg-type] urllib3.exceptions.MaxRetryError: HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /hunyuanvideo-community/HunyuanVideo/resolve/main/text_encoder/config.json (Caused by ConnectTimeoutError(, 'Connection to huggingface.co timed out. (connect timeout=10)')) During handling of the above exception, another exception occurred: Traceback (most recent call last): File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\file_download.py", line 1484, in _get_metadata_or_catch_error metadata = get_hf_file_metadata( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\utils\_validators.py", line 114, in _inner_fn return fn(*args, **kwargs) File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\file_download.py", line 1401, in get_hf_file_metadata r = _request_wrapper( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\file_download.py", line 285, in _request_wrapper response = _request_wrapper( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\file_download.py", line 308, in _request_wrapper response = get_session().request(method=method, url=url, **params) File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\requests\sessions.py", line 589, in request resp = self.send(prep, **send_kwargs) File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\requests\sessions.py", line 703, in send r = adapter.send(request, **kwargs) File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\utils\_http.py", line 96, in send return super().send(request, *args, **kwargs) File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\requests\adapters.py", line 507, in send raise ConnectTimeout(e, request=request) requests.exceptions.ConnectTimeout: (MaxRetryError("HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /hunyuanvideo-community/HunyuanVideo/resolve/main/text_encoder/config.json (Caused by ConnectTimeoutError(, 'Connection to huggingface.co timed out. (connect timeout=10)'))"), '(Request ID: b32a9bd0-fe3a-438c-adc0-8c7dadf2fd03)') The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\transformers\utils\hub.py", line 403, in cached_file resolved_file = hf_hub_download( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\utils\_validators.py", line 114, in _inner_fn return fn(*args, **kwargs) File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\file_download.py", line 961, in hf_hub_download return _hf_hub_download_to_cache_dir( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\file_download.py", line 1068, in _hf_hub_download_to_cache_dir _raise_on_head_call_error(head_call_error, force_download, local_files_only) File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\huggingface_hub\file_download.py", line 1599, in _raise_on_head_call_error raise LocalEntryNotFoundError( huggingface_hub.errors.LocalEntryNotFoundError: An error happened while trying to locate the file on the Hub and we cannot find the requested files in the local cache. Please check your connection and try again or make sure your Internet connection is on. The above exception was the direct cause of the following exception: Traceback (most recent call last): File "C:\Users\cain1\Downloads\framepack_cu126_torch26\webui\demo_gradio.py", line 49, in text_encoder = LlamaModel.from_pretrained("hunyuanvideo-community/HunyuanVideo", subfolder='text_encoder', torch_dtype=torch.float16).cpu() File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\transformers\modeling_utils.py", line 3613, in from_pretrained config, model_kwargs = cls.config_class.from_pretrained( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\transformers\configuration_utils.py", line 545, in from_pretrained config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs) File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\transformers\configuration_utils.py", line 574, in get_config_dict config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs) File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\transformers\configuration_utils.py", line 633, in _get_config_dict resolved_config_file = cached_file( File "C:\Users\cain1\Downloads\framepack_cu126_torch26\system\python\lib\site-packages\transformers\utils\hub.py", line 446, in cached_file raise EnvironmentError( OSError: We couldn't connect to 'https://huggingface.co' to load this file, couldn't find it in the cached files and it looks like hunyuanvideo-community/HunyuanVideo is not the path to a directory containing a file named text_encoder\config.json. Checkout your internet connection or see how to run the library in offline mode at 'https://huggingface.co/docs/transformers/installation#offline-mode'.

Jeremiah Everett

I started over and reran everything not as admin and still have the same results! any advice? Please help.

Dragonbait

how do you update, I can't find the update.bat file?

Rikard

i installed it, but when i run the bat file to download the models i just get the "The paging file is too small for this operation to complete" error.

Aitrepreneur

This error can be solved by Increasing the Page File size in windows, google it it's not hard to do, there are youtube tutorials that show how to do that

Rikard

i have increased the Page File size now, and it has downloaded all the 4 shards but upon trying to load them, it just seems to crash. there is no error message, it just immediately freezes the "loading checkpoint shards" at 0% and gives the "press any button to continue" prompt to close the window

Aitrepreneur

You need to increase it more and make sure you have enough spaces on your disk

Frunk

This was a lot of fun to play with but I found every video after 10 seconds just kinda froze out as it had kinda exhausted your instruction. I would recommend peeps try this out: https://github.com/colinurbs/FramePack-Studio Exactly the same (in fact you can install it so it uses the same files you already have) but with 2 important changes I found useful: 1. You can create a timestamped prompt - i.e. it gives the video generator a new prompt over time timeline you set for the video. i.e. first 5 secs show a forest scene. At 10 secs: have a deer dart across the scene 2. It lets you include LORAs... you naughty, naughty people... It is pretty trivial to install if you don't have clashing dependencies... just follow the install on the repo. I did delete the old FramePack install from here to avoid confusion. But all that side still works as before.

asdf

In run.exe Sage Attn is installed, so everything worked. However, checkpoint shards never loads beyond 0%, eventually leading to a "press any key to continue" notification. Web browser never launches. I have deleted and rerun everything about 6/7 times now and ot always gets hung up at that point. Can anyone help?

Thoritarian

I am running a 2070 Super on a laptop. I installed CUDA 12.6. I ran the V2 installer. I chose #3 and 'N'. I did not run upgrade.bat. I get the same error. I tried the above pip instructions. It gets further but now I get the following results. Suggestions? Thanks! C:\Users...\attn_qk_int8_per_block.py:18:23: error: 'tt.fp_to_fp' op operand #0 must be floating-point or ranked tensor of floating-point values, but got 'tensor<128x128xi8, #ttg.dot_op<{opIdx = 0, parent = #ttg.blocked<{sizePerThread = [4, 4], threadsPerWarp = [2, 16], warpsPerCTA = [8, 1], order = [1, 0]}>}>>' qk = tl.dot(q, k).to(tl.float32) * q_scale * k_scale ^

Nampot

Is this fixed yet?

Aitrepreneur

This error can be solved by Increasing the Page File size in windows, google it it's not hard to do, there are youtube tutorials that show how to do that

asdf

Page file size increase worked. Changed mine to min 3000 max 50000

Robbieddanger

I drag my pictures over, add my prompt and start the process. No preview comes up and the image is outputted as a png files.

Navyguy87

need some help i have AMD gpu is there anything i can you use for image to video

Navyguy87

System Manufacturer: Alienware System Model: Alienware Aurora Ryzen Edition R14 System Type: x64-based PC Processor(s): 1 Processor(s) Installed. [01]: AMD64 Family 25 Model 33 Stepping 2 AuthenticAMD ~3001 Mhz

Shivam Singhal

cant get it to produce uncensored stuff

Colossal Pictures

hello I am getting an error message when trying to install saying can not import from hugging face, any advice on this?

GlobalGrooveWORLD

Will there be a new studio version? The old one's a big loser.

Tonic69

I tried the new V2 on my 5090, but the file framepack_cu126_torch26.7z never gets created and there isn't a framepack folder after it all runs. But it says installation complete.

Joseph Daddona

So I saw here to extend my page. I did do that but when I tried to run.bat again it said all was done but it doesn't open up anything and I see no URL. So I deleted everything to start again. When I run your program it no longer makes the frame pack folder. Just a zip file that can't be extracted. I've tried restarting my comp and using a new download of your program. Any help?

Joseph Daddona

So I got everything to download, including the shaders, but no URL. It says press any key to continue, and it just closes down and doesn't open anything.

Marc

Had to tinker around a while because I had issues with my 2070 Super. My solution; 1. Get a special repo for FramePack: https://github.com/freely-boss/FramePack-nv20 2. Head there, open a terminal, install requirements.txt python -m pip install -r requirements.txt 3. Install all dependencies (finding the correct ones was a pain in the butt) python -m pip uninstall -y torch torchvision torchaudio python -m pip install torch==2.7.0+cu126 torchvision==0.22.0+cu126 torchaudio==2.7.0 --extra-index-url https://download.pytorch.org/whl/cu126 python -m pip install --force-reinstall --no-deps xformers 4. Run demo_gradio.py python demo_gradio.py Make sure to be on Python 3.10 (google or chatgpt will help you get there) so it works.

Rusty

This is great! Wondering if there are any workflows you have for adding lip synced audio to videos. 🤔

Ng Luke

got error ... torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 14.34 GiB. GPU 0 has a total capacity of 11.64 GiB of which 3.04 GiB is free. Process 1452659 has 8.59 GiB memory in use. Of the allocated memory 6.13 GiB is allocated by PyTorch, and 2.12 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation.